Skip to content

keyError: 'layout' #3

@lpdswing

Description

@lpdswing

marker_pdf_gradio | Watching: '/workspace/gradio' '/workspace/gradio'
marker_pdf_gradio |
marker_pdf_gradio | Some weights of SuryaModel were not initialized from the model checkpoint at /root/.cache/datalab/models/text_recognition/2025_08_04 and are newly initialized: ['multi_output_projections.0.fc1.bias', 'multi_output_projections.0.fc1.weight', 'multi_output_projections.0.fc2.bias', 'multi_output_projections.0.fc2.weight', 'multi_output_projections.1.fc1.bias', 'multi_output_projections.1.fc1.weight', 'multi_output_projections.1.fc2.bias', 'multi_output_projections.1.fc2.weight', 'multi_output_projections.2.fc1.bias', 'multi_output_projections.2.fc1.weight', 'multi_output_projections.2.fc2.bias', 'multi_output_projections.2.fc2.weight', 'multi_output_projections.3.fc1.bias', 'multi_output_projections.3.fc1.weight', 'multi_output_projections.3.fc2.bias', 'multi_output_projections.3.fc2.weight']
marker_pdf_gradio | You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
marker_pdf_gradio | Traceback (most recent call last):
marker_pdf_gradio | File "/workspace/gradio/gradio_app.py", line 100, in
marker_pdf_gradio | model_dict = load_models()
marker_pdf_gradio | ^^^^^^^^^^^^^
marker_pdf_gradio | File "/workspace/gradio/gradio_app.py", line 47, in load_models
marker_pdf_gradio | return create_model_dict()
marker_pdf_gradio | ^^^^^^^^^^^^^^^^^^^
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/marker/models.py", line 12, in create_model_dict
marker_pdf_gradio | foundation_predictor = FoundationPredictor(device=device, dtype=dtype)
marker_pdf_gradio | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/surya/foundation/init.py", line 96, in init
marker_pdf_gradio | super().init(checkpoint, device, dtype)
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/surya/common/predictor.py", line 29, in init
marker_pdf_gradio | self.processor = loader.processor()
marker_pdf_gradio | ^^^^^^^^^^^^^^^^^^
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/surya/foundation/loader.py", line 64, in processor
marker_pdf_gradio | ocr_tokenizer = SuryaOCRTokenizer(
marker_pdf_gradio | ^^^^^^^^^^^^^^^^^^
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/surya/common/surya/processor/tokenizer.py", line 252, in init
marker_pdf_gradio | self.ocr_tokenizer = InnerOCRTokenizer(
marker_pdf_gradio | ^^^^^^^^^^^^^^^^^^
marker_pdf_gradio | File "/opt/conda/lib/python3.11/site-packages/surya/common/surya/processor/tokenizer.py", line 54, in init
marker_pdf_gradio | self.LAYOUT_TAG_PATTERN = create_token_regex(special_tokens["layout"])
marker_pdf_gradio | ~~~~~~~~~~~~~~^^^^^^^^^^
marker_pdf_gradio | KeyError: 'layout'
marker_pdf_gradio exited with code 0

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions